Regulating Artificial Intelligence in Brazil

On May 25, 2023, the Center for Human Rights and Global Justice’s Technology & Human Rights team hosted an event entitled Regulating Artificial Intelligence: The Brazilian Approach, in the fourteenth episode of the “Transformer States” interview series on digital government and human rights. This in-depth conversation with Professor Mariana Valente, a member of the Commission of Jurists created by the Brazilian Senate to work on a draft bill to regulate artificial intelligence, raised timely questions about the specificities of ongoing regulatory efforts in Brazil. These developments in Brazil may have significant global implications, potentially inspiring other more creative, rights-based, and socio-economically grounded regulation of emerging technologies in the Global South.

In recent years, numerous initiatives to regulate and govern Artificial Intelligence (AI) systems have arisen in Brazil. First, there was the Brazilian Strategy for Artificial Intelligence (EBIA), launched in 2021. Second, legislation known as Bill 21/20, which sought to specifically regulate AI, was approved by the House of Representatives in 2021. And in 2022, a Commission of Jurists was appointed by the Senate to draft a substitute bill on AI. This latter initiative holds significant promise. While the EBIA and Bill 21/20 were heavily criticized for the limited value given to public input in comparison to the available participatory and multi-stakeholder mechanisms, the Commission of Jurists took specific precautions to be more open to public input. Their proposed alternative draft legislation, which is grounded in Brazil’s socio-economic realities and legal tradition, may inspire further legal regulation of AI, especially for the Global South, considering Brazil’s position in other discussions related to internet and technology governance.

Bill 21/20 was the first bill directed specifically at AI. But this was a very minimal bill; it effectively established that regulating AI should be the exception. It was also based on a decentralized model, meaning that each economic sector would regulate its own applications of AI: for example, the federal agency dedicated to regulating the healthcare sector would regulate AI applications in that sector. There were no specific obligations or sanctions for the companies developing or employing AI, and there were some guidelines for the government on how it should promote the development of AI. Overall, the bill was very friendly to the private sector’s preference for the most minimal regulation possible. The bill was quickly approved in the House of Representatives, without public hearings or much public attention.

It is important to note that this bill does not exist in isolation. There is other legislation that applies to AI in the country, such as consumer law and data protection law, as well as the Marco Civil da Internet (Brazilian Civil Rights Framework for the Internet). These existing laws have been leveraged by civil society to protect people from AI harms. For example, Instituto Brasileiro de Defesa do Consumidor (IDEC), a consumer rights organization, successfully brought a public civil action using consumer protection legislation against Via Quatro, a private company responsible for the subway line 4-Yellow of Sao Paulo. The company was fined R$500,000 for collecting and processing individuals’ biometric data for advertising purposes without informed consent.

But, given that Bill 21/20 sought to specifically address the regulation of AI, academics and NGOs raised concerns that it would reduce the legal protections afforded in Brazil: it “gravely undermines the exercise of fundamental rights such as data protection, freedom of expression and equality” and “fails to address the risks of AI, while at the same time facilitating a laissez-faire approach for the public and private sectors to develop, commercialize and operate systems that are far from trustworthy and human-centric (…) Brazil risks becoming a playground for irresponsible agents to attempt against rights and freedoms without fearing for liability for their acts.”

As a result, the Senate decided that instead of voting on Bill 21/20, they would create a Commission of Jurists to propose a new bill.

The Commission of Jurists and the new bill

The Commission of Jurists was established in April 2022 and delivered its final report in December 2022. Even if the establishment of the Commission was considered a positive development, it was not exempt from criticism from civil society, for the lack of racial and regional diversity of the Commission’s membership, as well as the need for different areas of knowledge to contribute to the debate. This criticism comes from a reflection of the socio-economic realities of Brazil, which is one of the most unequal countries in the world, and those inequalities are intersectional, considering race, gender, income, territorial origin. Therefore, AI applications will have different effects on different segments of the population. This is already clear from the use of facial recognition in public security: more than 90% of the individuals arrested using this technology were Black. Another example is the use of an algorithm to evaluate requests for emergency aid amid the pandemic, where many vulnerable people had their benefits denied based on incorrect data.

During its mandate, the Commission of Jurists held public hearings, invited specialists from different areas of knowledge, and developed a public consultation mechanism allowing for written proposals. Following this process, the new proposed bill had several elements that were very different from Bill 21/20. First, the new bill borrows from the EU’s AI Act by adopting a risk-based approach: obligations are distinguished according to the risks they pose. However, the new bill, following the Brazilian tradition of structuring regulation from the perspective of individual and collective rights, merges the European risk-based approach with a rights-based approach. The bill confers individual and collective rights that apply in relation to all AI systems, independent of the level of risk they pose.

Secondly, the new bill includes some additional obligations for the public sector, considering its differential impact on people’s rights. For example, there is a ban on the treatment of racial information, and provisions on public participation in decisions regarding the adoption of these systems. Importantly, though the Commission discussed the inclusion of a complete ban on facial recognition technologies in public spaces for public security, this proposal was not included: instead, the bill included a moratorium, establishing that a law must be approved regulating this use.

What the future holds for AI regulation in Brazil

After the Commission submitted its report, in May 2023 the president of the Senate presented a new bill for AI regulation replicating the Commission’s proposal. On 16th August 2023, the Senate established a temporary internal commission to discuss the different proposals for AI regulation that have been presented in the Senate to date.

It is difficult to predict what will happen following the end of the internal commission’s work, as political decisions will shape the next developments. However, what is important to have in mind is the progress that the discussion has reached so far, from an initial bill that was very minimal in scope, and supported the idea of minimal regulation, to one that is much more protective of individual and collective rights and considerate of Brazil’s particular socio-economic realities. Brazil has played an important progressive role historically in global discussions on the regulation of emerging technologies, for example with the discussions of its Marco Civil da Internet. As Mariana Valente put it, “Brazil has had in the past a very strong tradition of creative legislation for regulating technologies.” The Commission of Jurists’ proposal repositions Brazil in such a role.

September 28, 2023. Marina Garrote, LLM program, NYU School of Law whose research interests lie at the intersection of digital rights and social justice. Marina holds a bachelor and master’s degree from Universidade de São Paulo and previously worked at Data Privacy Brazil, a civil society association dedicated to public interest research on digital rights.